Longitudinally changed diet quality scores and their association with type 2 diabetes mellitus and cardiovascular diseases in the epic-potsdam study

ABSTRACT Association analyses between longitudinal changes in diet quality scores (DQIs) and cardiometabolic risk remain scarce. Hence, we aimed to investigate how changes in two DQIs are

associated with incident type 2 diabetes (T2D), myocardial infarction (MI) and stroke in the EPIC-Potsdam study. Changes in the Mediterranean Pyramid Score (MedPyr) and Healthy Diet Score

(HDS) over 7 years from baseline (1994–1998) to follow-up 3 (2001–2005) were investigated in 23,548 middle-aged participants. Adjusted Cox Proportional Hazards Regression models were applied

to investigate associations between changes in MedPyr and HDS and chronic disease incidence. More than 60% of the participants increased both DQIs more than 5%. Within a median follow-up

time of 5 years 568 cases of T2D, 171 of MI, 189 of stroke were verified. An increased compared to stable MedPyr was associated with lower T2D risk (HR 0.74; 95% CI 0.59–0.92), while a

decreased MedPyr was associated with higher stroke risk (HR 1.67; 95% CI 1.02–2.72). A decreased compared to stable HDS was associated with higher stroke risk (HR 1.80; 95% CI 1.02–3.20).

The findings contribute further evidence on advantages of changing dietary intake towards a Mediterranean Diet. Although baseline HDS adherence was associated with T2D and stroke risk,

longitudinal changes in HDS were only significantly associated with stroke risk. SIMILAR CONTENT BEING VIEWED BY OTHERS HEALTHY FOOD DIVERSITY AND THE RISK OF MAJOR CHRONIC DISEASES IN THE

EPIC-POTSDAM STUDY Article Open access 19 November 2024 SHORT-TERM IMPROVEMENTS IN DIET QUALITY IN PEOPLE NEWLY DIAGNOSED WITH TYPE 2 DIABETES ARE ASSOCIATED WITH SMOKING STATUS, PHYSICAL

ACTIVITY AND BODY MASS INDEX: THE 3D CASE SERIES STUDY Article Open access 13 July 2020 DIETARY HABITS AND COMPLIANCE WITH DIETARY GUIDELINES IN PATIENTS WITH ESTABLISHED CARDIOVASCULAR

DISEASE Article 27 May 2024 INTRODUCTION Besides the investigation of individual foods and nutrients, the study of dietary patterns acts as a complementary strategy to disentangle the

complexity of human diet. It has the advantage to consider the interrelation of food intake and to represent the cumulative exposure to different diet components, potentially leading to

stronger effects on health than any single component1. Numerous prospective studies have evaluated associations between dietary patterns and cardiometabolic diseases2,3. Regarding a priori

(hypothesis-based) dietary patterns (DP), evidence is strongest for the Mediterranean diet2,4,5, but also the Alternative Healthy Eating Index (AHEI) and Dietary Approaches to Stop

Hypertension (DASH) were associated with a reduced risk of type 2 diabetes (T2D) and cardiovascular diseases (CVD)2,3,6. Noteworthy, studies conducted outside the Mediterranean region, thus

in study populations which generally follow different dietary habits, e.g. European Prospective Investigation into Cancer and Nutrition (EPIC)-Potsdam study among others, also observed lower

cardiometabolic risk for participants following a Mediterranean diet7,8,9. Beneficial effects of the Mediterranean diet are strongly supported by the Prevención con Dieta Mediterránea

(PREDIMED) randomized trial, which showed lower CVD risk with this diet, being either supplemented with extra virgin olive oil or with nuts10. Previous studies mostly used dietary

information collected from only one time point. However, using dietary information repeatedly collected during follow-up could be used to reduce measurement error due to intra-individual

variation11. Furthermore, in terms of supporting dietary guidelines, it is of particular interest to investigate, if an actual change in dietary intake is associated with the incidence of

chronic diseases. To our knowledge, few studies have already investigated the effect of changes in DPs over time on chronic disease risk: A decrease in diet quality assessed by the AHEI was

related to higher T2D risk, while an increase in diet quality was related to lower risk in three US cohorts12. Similarly, longitudinal increases in the diet quality scores alternative

Mediterranean diet (aMED), AHEI and DASH were associated with a lower CVD risk13. In European cohorts, most investigations on longitudinal changes of DP were restricted to associations with

intermediate cardiometabolic risk factors14,15,16 and one study used diet data from repeated assessments, but calculated the cumulative average and its association with cardiovascular

disease risk9. To close the knowledge gap on a potential association of changes in dietary behavior with chronic disease endpoints in European cohorts, we aimed to investigate the

longitudinal change in diet quality scores and its association with incident T2D, MI and stroke in the EPIC-Potsdam study, a prospective cohort with repeated dietary assessments and

follow-up on cardiometabolic diseases. SUBJECTS AND METHODS STUDY POPULATION Participants were recruited between 1994 and 1998 from the general population living in Potsdam and surrounding

areas, with 16,644 women aged 35 to 65 years and 10,904 men aged 40 to 65 years participating17. At baseline, participants were examined by trained personnel for anthropometric assessment

and answered questions on their prevalent diseases, diet and lifestyle via interviewer-based and self-administered questionnaires18. Informed consent was obtained from all participants and

the study was approved by the Ethical Committee of the State of Brandenburg, Germany and follows the recommendations of the Declaration of Helsinki. During follow-up, the participants were

contacted again every 2–3 years to identify incident cases of the respective diseases of interest19, and were asked to fill in a questionnaire on their habitual dietary intake at the third

follow-up (FUP3) between 2001 and 200520. For this analysis, participants with information on their vital status until the fifth follow-up (FUP5) (2007–2009) were included. The follow-up

rates ranged between 90 and 96% for each follow-up round. At FUP3 26,345 participants were still active (Supplementary Fig. 1). From these, we excluded participants with incomplete

questionnaire data (n = 2147). Further exclusions on missing diet (n = 10) and covariates (n = 640) were done, resulting in a study population of n = 23,548 participants. We were interested

in the change in diet from baseline to FUP3, and the subsequent occurrence of incident T2D and CVD until FUP5. Hence, we excluded participants who reported to be diagnosed with T2D, MI or

stroke at baseline and those who developed T2D (n = 2155), MI (n = 597) or stroke (n = 461) until FUP3 for the disease-specific association analyses (Supplementary Fig. 1). DIETARY

ASSESSMENT The habitual intake of diet was assessed with a semi-quantitative FFQ at baseline (FFQ0) comprising 149 food items. The frequency of intake was inquired with 10 categories ranging

from “I don’t eat it” to “5 times per day or more”. Additional pictures of portion sizes supported participants to estimate their consumed quantity. The respective standard portion size was

adopted from a list of food items, which was developed in the framework of the MONItoring of trends and determinants of CArdiovascular disease (MONICA) project in Augsburg, Germany21. With

regard to selected food groups prone to misreporting due to e.g. seasonal variance, summation questions on those food groups were used to determine a correction factor for the calculation of

the overall habitual intake in g/day. At FUP3, participants were asked again to complete an updated FFQ (FFQ1), which comprised a shortened list of 105 food items. The update of the FFQ was

conducted systematically to restrict the number of inquired food items to those explaining maximum variance in food groups and selected nutrients, and to include new food items to consider

trends in food intake over time20. Portion sizes were provided by common household measures and the frequency of consumption was asked in 5 to 10 response categories, depending on the

inquired food items. Systematic differences between the two FFQs were investigated and corrected. Details are given in the Supplementary Methods “Systematic investigation of differences

between FFQ0 and FFQ1” and Supplementary Table 1. CHOICE OF DIET QUALITY SCORES AND CALCULATION OF THEIR LONGITUDINAL CHANGES In our previous investigation of the association between

baseline diet quality scores with the incidence of T2D, myocardial infarction (MI) and stroke in the EPIC-Potsdam study, we focused on the Mediterranean diet, using a scoring originally

derived by Trichopoulou et al. (tMDS)22 and the Mediterranean Pyramid (MedPyr) score, and the Nordic diet7. In the present study, investigating longitudinal changes in adherence to diet

quality scores, we considered only the MedPyr to reflect a Mediterranean diet, as it was more clearly associated with lower incident T2D in EPIC-Potsdam in the former analysis7. Instead of

the Nordic diet, which was not associated with the chronic diseases in our study population7, we decided to investigate a newly developed diet quality score for healthy eating (Healthy Diet

Score—HDS)23. This considers the current German dietary guidelines and latest evidence of beneficial and detrimental associations of food groups with the chronic diseases of interest. The

diet scores have been investigated according to their relative validity and reliability in a validation study (n = 134) embedded in EPIC-Potsdam23,24. For both scores, reliability and

relative validity were moderate. The MedPyr showed a mean difference of 0.03 score points and a correlation of r = 0.62 based on two similar FFQs applied one year apart, while the HDS showed

a mean difference of 0.03 score points and a correlation of r = 0.53. Relative validity was lower but still moderate: the FFQ derived data were compared to the mean of 12 24-h dietary

recalls within one year (MedPyr: mean difference − 0.31 and r = 0.45; HDS: mean difference 0.29 and r = 0.43). Details of the construction of both scores were described in previous

publications. Briefly, the MedPyr comprises 15 food groups reflecting guidelines of the Mediterranean pyramid7,9 (Supplementary Table 2), while the HDS comprises 10 food groups, which were

selected based on German dietary guidelines complemented by latest evidence of their beneficial and detrimental potential for the development of chronic diseases23 (Supplementary Table 3).

While the MedPyr theoretically ranges from 0 to 15 points, the HDS ranges from 0 to 10 points. The relative change in diet quality scores between baseline and FUP3 (in percent) was

calculated as the absolute difference between the scores from FUP3 and baseline, divided by the baseline score and multiplied by 100 (Formula 1). $${diet change}_{perc}=\frac{{diet quality

score}_{FUP3}-{diet quality score}_{Baseline}}{{diet quality score}_{Baseline}}*100$$ (1) A change of 5% or less in the two diet quality scores was categorized as “stable”, while more than

5% decrease in the diet quality scores was categorized as “decrease” and more than 5% increase in the scores as “increase”, partly comparable to the approaches in previous U.S. based

cohorts12,25. An overview of the food groups at baseline and at FUP3 across the change categories of both diet quality indices was provided in Supplementary Tables 4, 5. COVARIATE ASSESSMENT

Baseline and FUP3 information on socio-demographic covariates like age, sex and occupation, lifestyle factors like alcohol consumption and smoking behavior, as well as self-reported

hypertension were obtained by self-administered questionnaires and an interview (at baseline), except for physical activity (lastly updated information at FUP2) and education (only baseline

information). Categories of highest educational level were “currently in training/no certificate or skill”, “professional school (vocational training)”, and “college or higher education”.

Occupational status was a variable with the three categories “employed”, “unemployed” and “retired”. Smoking status was categorized as “never smoker”, “former smoker” and “current smoker”.

If the updated information on smoking behavior was not reported at FUP3, the missing information was imputed based on the reported information at a later follow-up. Physical activity was

obtained as combined information of leisure time sports activities and biking, both in hours/week. Lifetime alcohol consumption combined the information on alcohol intake during life course

(at the age 20, 30 and 40) and the intake estimate from baseline and FUP3, respectively26. The anthropometrics of the participants were assessed by trained staff at baseline, who followed a

standard protocol with a strict quality control27. Body weight was measured by electronic digital scales, accurate to 0.1 kg and heights were measured to the nearest 0.1 cm using a flexible

anthropometer28. At FUP3 body weight was self-reported by the participants. The Body-Mass-Index (BMI) was calculated by dividing the body weight in kg by the square of height in m. OUTCOME

ASCERTAINMENT Incident T2D, MI and stroke were ascertained by self-reports of the disease, disease-relevant medication or diagnosis as a reason for change in diet at the follow-ups. Further

potential cases were obtained from medical databases, death certificates or other sources like physicians or clinics. All potential cases had to be verified by the primary care physician,

including the correct diagnosis, the diagnostic test and exact date of diagnosis19. The respective International Classification of Diseases (ICD-10) codes I21 for MI, I60, I61, I63 and I64

for stroke and E11 for T2D were used. STATISTICAL ANALYSES Characteristics of the participants are described across the diet quality score change categories as mean ± standard deviation, if

normally distributed and as median (interquartile range), if the assumption of normal distribution was not fulfilled. Categorical variables are presented as relative proportions. Since the

HDS has not been investigated according to its association with chronic disease risk before, an analysis was conducted, where the association of baseline adherence to the HDS (in quintiles)

as exposure with the incidence of all three outcomes until the FUP5 (mean follow-up time: 11 years) was investigated. For this, a Cox Proportional Hazards Regression model was set up, where

the time between age at recruitment and age at exit was used as underlying time scale and the model was stratified by integers of age. To account for potential covariates, the model was

adjusted for sex, education, occupational status, vitamin supplementation, self-reported hypertension, smoking status, physical activity, lifetime alcohol consumption, total energy intake

and BMI, all covariates respectively assessed at baseline to achieve comparability to a previous investigation on the association between adherence to other diet quality scores and

cardiometabolic diseases in our study population7. Associations between the change of both respective diet quality scores as exposure and the incident chronic diseases as outcome were

calculated as multivariable-adjusted hazard ratios (HR) with 95% confidence intervals by Cox Proportional Hazards Regression models. The change in diet quality scores was modeled as (a)

continuous exposure per 20%-increase (corresponds to 3-points increase for MedPyr; 2 points increase for HDS) and (b) the percentual change as categorical exposure, setting “stable” as

reference category to account for potential trends in decreases or increases. The dependent time variable was the time between age at FUP3 and age at exit (either date of diagnosis, date of

death or censoring at the end of FUP5). The models were stratified by integers of age to be less sensitive to violations of the HR by different age-dependent risk at analysis baseline. Model

1 was adjusted for sex, the respective diet quality score and education at baseline, while baseline and updated FUP3 information was used for occupational status, self-report of

hypertension, smoking status and lifetime alcohol consumption, total energy intake, and physical activity (taken from FUP2, where the information was available), respectively. Model 2 was

additionally adjusted for baseline BMI and BMI at FUP3 to account for potential confounding on the diet change-disease-association. Schönfeld residuals were calculated to check the

proportional hazards assumptions. Non-linear associations between the change in diet quality scores and the outcomes of interest were evaluated with restricted cubic splines (RCS) and a Wald

χ2 test. To assess potential effect measure modification of the association between the change in diet quality and the respective outcomes by sex29,30 or by baseline diet quality score,

stratified analyses were conducted. Furthermore, multiplicative interaction terms with the respective diet quality scores were added to Model 2 and tested in a Wald test for significant

interaction. To account for potential misreporting, participants with implausible energy intake (women: < 500 kcal/day, > 3500 kcal/day; men: < 800 kcal/day, > 4200 kcal/day) at

baseline (n = 410) and at FUP3 (n = 265) were excluded in a sensitivity analysis. Those participants, who reported a change in diet at FUP3 due to weight gain (n = 5536), high blood pressure

(n = 143), dyslipidemia (n = 596) or diabetes (n = 766) within the previous 2 years, were excluded in further sensitivity analyses. To investigate, if potential beneficial effects of the

MedPyr score changes on the outcomes were driven by moderate alcohol consumption, this component was excluded in a sensitivity analysis. All analyses were performed with SAS Software version

9.4 (SAS Institute Cary, NC). ETHICS APPROVAL This study was performed in line with the principles of the Declaration of Helsinki. All study procedures were approved by the Ethical

Committee of the State of Brandenburg. CONSENT TO PARTICIPATE Participants agreed to participate in the EPIC-Potsdam study by signing the informed consent form. RESULTS Table 1 shows that

more than 60% of the 23,548 included participants increased adherence to both diet quality scores by more than 5% between baseline and FUP3 (median time span: 6.6 years), while approximately

similar proportions of participants decreased their diet quality scores or remained relatively stable. Participants in the increase-category for both diet quality scores were less likely to

be men, started with a lower diet quality score at baseline and lower total energy intake. No differences across the change categories were observed for baseline age, BMI or physical

activity. While alcohol consumption was considerably lower in participants, who increased adherence to MedPyr, the consumption was not different between participants, who increased adherence

to HDS compared to those who remained stable. The increase in both diet quality scores was in line with the higher rate of self-reported change in diet in the previous 2 years at FUP3.

ASSOCIATIONS OF CHANGES IN HDS WITH CHRONIC DISEASES Higher adherence to the baseline HDS resulted in both significant lower T2D risk (HRQ5 vs Q1 = 0.82; 95% CI 0.69–0.97; HR per 1 SD = 

0.90; 95% CI 0.85–0.96) and lower stroke risk (HRQ5 vs Q1 = 0.57; 95% CI 0.40–0.82; HR per 1 SD = 0.83; 95% CI 0.73–0.94) over a mean follow-up of 11 years (Supplementary Table 6). The

analysis of the change in HDS did not result in significant associations for the three chronic disease outcomes when the change was modelled continuously (Table 2). The RCS analysis did not

indicate any deviation from linearity (Supplementary Fig. 2A,C,E). Still, when comparing categories of change, participants who decreased their HDS had higher risk of stroke (HR 1.80; 95% CI

1.02–3.20) compared to participants with stable HDS (Table 2). Interaction analyses did not support that associations differed between men and women (all p for interaction > 0.05) and we

did not observe any significant associations in stratified analyses by sex (Supplementary Table 7). The stratified analyses by baseline HDS tertiles did not indicate significant

associations for the 20% increase in HDS with incident T2D and stroke, but a significant lower MI risk (HR 0.66; 95% CI 0.49–0.91) in those participants starting in the “medium” HDS tertile

at baseline (Supplementary Table 8). Still, none of the interaction terms were statistically significant (all p for interaction > 0.05). Since the RCS analysis indicated non-linear

associations for stroke risk in the “low” and “high” HDS baseline tertile (Supplementary Fig. 3), additional analyses for the comparison of change categories were conducted. Starting with

“low” HDS tertile at baseline, an increase in HDS compared to the stable reference pointed towards a lower stroke risk (HR 0.48; 95% CI 0.21–1.13). Starting in the “high” HDS tertile at

baseline and decrease the HDS compared to the stable category pointed towards a higher stroke risk (HR 2.29; 95% CI 0.93–5.66) (data not shown). The sensitivity analyses on the exclusion of

participants with implausible energy intake or those, who reported a change in diet due to weight gain, blood pressure, dyslipidemia or diabetes diagnosis did not indicate altered

associations between a longitudinal HDS change and the three chronic disease outcomes compared to the main results (Supplementary Table 9). ASSOCIATIONS OF CHANGES IN MEDPYR WITH CHRONIC

DISEASES The RCS analysis did not show any deviation from linearity for the association between the change in MedPyr and the three outcomes of interest (Supplementary Fig. 2B,D,F). A 20%

increase in MedPyr, modelled continuously, pointed towards a lower T2D risk (HR 0.94; 95% CI 0.85–1.03) over a median follow-up time of 4.7 years (Table 2). When comparing participants in

the increase-category to the stable-category as reference, they had a lower T2D risk in Model 1 and this association was not altered by the adjustment for BMI in Model 2 (HR 0.74; 95% CI

0.59–0.92). Participants in the decrease-category had similar risk for T2D in comparison to participants in the stable-category. Longitudinal changes in the MedPyr were not significantly

associated with the incidence of MI, neither when modelling changes on a continuous scale, nor when using categories. For the association of a longitudinal MedPyr change with stroke, the

analysis per 20% increase in MedPyr pointed towards an inverse association (HR 0.89; 95% CI 0.75–1.05), and the comparison of change categories indicated for participants in the

decrease-category compared to the stable reference a significant higher stroke risk (HR 1.67; 95% CI 1.02–2.72). The stratification of the analysis by sex did not indicate significant

differences in the associations of MedPyr change with MI and stroke, but pointed towards a lower T2D risk for the 20% increase in MedPyr in women (HR 0.89; 95% CI 0.77–1.04) (Supplementary

Table 7). Still, none of the interaction terms reached statistical significance. We next evaluated if associations of changes in MedPyr with cardiometabolic endpoints depend on baseline

adherence to this diet. The interaction terms did not reach statistical significance (Supplementary Table 8). Still noteworthy, the stratification by baseline MedPyr tertiles pointed towards

a lower risk for all three outcomes per 20% increase in MedPyr in those participants, who started in the “medium” MedPyr tertile at baseline. RCS analysis indicated a non-linear association

with incident T2D in the “low” baseline MedPyr tertile (Supplementary Fig. 4). Here, the comparison of change categories pointed towards a lower T2D risk (HR 0.78; 95% CI 0.48–1.28) for an

increase in MedPyr compared to the stable category and towards a higher T2D risk (HR 1.47; 95% CI 0.78–2.78) for those, who decreased MedPyr compared to the stable reference (data not

shown). The RCS analysis did not indicate deviations from linearity across the baseline MedPyr tertiles for associations with MI and stroke (Supplementary Fig. 4). In a sensitivity analysis,

the removal of the alcohol component from MedPyr resulted in stronger inverse associations of MedPyr with incident T2D (HR per 20% increase = 0.90; 95% CI 0.81–1.00) and stroke (HR per 20%

increase = 0.83; 95% CI 0.70–1.00) (Supplementary Table 9). The exclusion of participants with implausible total energy intake, and the exclusion of those, who reported either a change in

diet due to weight gain, high blood pressure, dyslipidemia or diabetes, did not change the associations with T2D, MI and stroke compared to the main analysis (Supplementary Table 9).

DISCUSSION Repeated assessments of diet in the EPIC-Potsdam study allowed to quantify longitudinal change in diet quality scores adherence over a period of approximately 7 years and we

observed that participants, who increased their MedPyr by more than 5% had a significantly lower risk of T2D compared to participants with no changes in MedPyr. This inverse association

appeared to be more pronounced in women and those participants starting with a low and medium adherence to MedPyr at baseline. A decrease in MedPyr adherence by more than 5% was associated

with a higher risk for stroke, while changes in MedPyr score adherence were not associated with MI risk. A higher adherence to the HDS at baseline was associated with a lower risk for T2D

and stroke. Longitudinal changes of the HDS were not clearly associated with T2D and MI risk, but a decrease in HDS adherence by more than 5% was associated with a higher stroke risk in our

study population. So far, we could only identify two studies that have investigated longitudinal changes in diet quality scores in relation to cardiometabolic disease risk: In the U.S., one

study investigated the association between AHEI changes and T2D risk and concluded for a > 10% increase in AHEI over 4 years a lower T2D risk12, but no study has yet investigated the

association with a change in a diet quality score reflecting Mediterranean diet with regards to T2D risk. However, the association of changes in several dietary quality scores, including

AHEI, aMED and DASH with CVD risk were investigated in three U.S. cohorts13. The investigators concluded no significant association for an increase in aMED with total CVD risk, but a lower

risk for stroke. However, a comparison to our results is limited due to the different composition and calculation of aMED and MedPyr7,13: While the MedPyr evaluates absolute intakes of 15

components of a typical healthful Mediterranean diet, the aMED is restricted to the evaluation of relative intakes of 9 components, leaving out dairy products, eggs, white meat, sweets and

potatoes, and considering the ratio of monounsaturated to saturated fatty acids instead of the consumption of olive oil. A study in England (EPIC-Norfolk) investigated different

Mediterranean diet scores and their associations with CVD outcomes by using repeated dietary data and showed inverse associations between diet scores which considered absolute intakes (e.g.

MedPyr) and CVD, ischaemic heart disease and stroke9. However, comparability to our results is constrained, because cumulative averages were calculated instead of changes in diet quality

scores. Our results on changes in MedPyr and risk of T2D extent our previous investigation on adherence to baseline MedPyr7, where a higher adherence to a Mediterranean diet was related to

lower risk of T2D. Our current analyses support that changes of dietary intake towards a Mediterranean diet over time can reduce T2D risk. Contrary to previous reported attenuated

associations with T2D after exclusion of the alcohol component from a Mediterranean diet score in the EPIC-InterAct consortium31, our results indicated an enhanced lower risk of T2D. Since

the participants in our study have not changed their alcohol consumption over time, the exclusion of this component could in our particular case enhance precision of the model which could

explain the result of this sensitivity analysis. Although adherence to baseline MedPyr was not associated with stroke risk in our cohort7, a decrease in MedPyr over time, in comparison to

participants with a stable MedPyr, pointed towards a higher risk for stroke. This underpins the added value of this change analysis. Interestingly, the inverse association of MedPyr with

stroke was more pronounced after exclusion of the alcohol component. This is unexpected, given that a systematic review including 27 cohorts with 1.4 million participants concluded a

J-shaped association for alcohol intake and stroke morbidity32. A recently published investigation of 8 different diet quality indices with major chronic diseases in three U.S. cohorts

concluded for 6 commonly used indices reflecting dietary guidelines or regional dietary habits, e.g. AHEI, DASH or aMED, significant disease risk reductions33. Comparable to an index like

the AHEI, the components for the HDS in our investigation were also selected based on the national dietary guidelines in Germany and evidence from single food associations with T2D, CVD and

cancer23. Similarly, we observed significant risk reductions for T2D and stroke in our study population, comparing extreme quintiles of baseline adherence to the HDS. Wang et al. also

investigated two empirically-derived mechanism-based indices using components with hyperinsulinemic and inflammatory properties34 and showed the strongest associations with disease risk

reduction in the three investigated U.S. populations33. Although our analysis was restricted to only two guideline-based diet quality scores, we contributed new insights into the

associations between longitudinal changes of these with the T2D, MI and stroke risk. Interestingly, although only a decrease in HDS adherence was associated with a significant higher stroke

risk in the main analysis, the stratified analysis by baseline HDS tertile pointed towards a reduction in stroke risk when participants increased HDS adherence at least in the “low” HDS

tertile. Similarly, Sotos-Prieto et al. observed for 4-year changes in aMED and DASH a reduced risk for stroke13. Several methodological constraints potentially explained the null

associations between changes in HDS with T2D and MI risk in our study: One aspect in evaluating dietary changes in cohort studies is random measurement error, which may result in

misclassification of individuals (e.g. false assignment to categories of decreased or increased diet quality score) and would most likely tend to attenuate associations. Different degrees of

measurement error for different diet quality scores may be an explanation for the null associations between the HDS and chronic disease outcomes in comparison to the MedPyr. This notion is

supported by differences in reliability measures, which we observed in a sub study in EPIC-Potsdam using repeated FFQs, filled out by participants one year apart. For the HDS, reliability

was moderate (r = 0.53) but lower than what we have observed for the MedPyr (r = 0.62)23,24. A reason could be the low reliability of some components, which are part of the HDS, but not of

the MedPyr. So, for example, the reliability of the aggregated food group “vegetable oils”, included in the HDS, was very low (r = 0.07), while olive oil, as component of the MedPyr, had a

moderate reliability (r = 0.54). A more general limitation is the relatively short follow-up time of 4.4 years in our study, because we only considered incident cases occurring after the

actual change in the two diet quality scores (after FUP3). The small number of incident cases may have limited statistical power, in particular for stratified and interaction analyses, and

resulted in relative imprecise estimates. However, a strength of the outcome ascertainment in our study was the very low misclassification rate of cases (T2D: 0.25%; MI: 0.15%, Stroke: 0.07)

and the two-step approach of using different sources like self-report, medication or reasons to change the diet to firstly identify potential disease cases (Supplementary Table 10), but a

mandatory second step of verification by the treating physician. The few non-verified cases were excluded from this analysis, but we would not expect a major change in results except for a

slight improvement in precision of the estimates due to a higher statistical power. We could not rule out that differences in the structure of the two applied FFQs at baseline and FUP3 had

an impact on the calculated longitudinal change in diet quality scores. However, an important initial step of our analysis was the evaluation of dietary data from a comparison study (n = 

512) to identify systematic differences between the two FFQs and to correct food groups for such differences. Up to now, consequences of changes in the structure of dietary assessment

instruments in longitudinal studies to capture certain trends of dietary intake, and its impact on disease-associations, have not been addressed systematically, but should get more attention

in future studies. As strengths of our analyses, we were able to consider dietary data not only from one time point, but additionally from a second time point in a large sample of more than

23,000 participants. Even more time points could have enabled the modeling of the exposure as time-varying change in diet score, as it was conducted in two U.S. studies12,13. This was not

possible in our analysis. Similar to another U.S. study, which investigated how change in DASH diet was associated with cardiovascular risk factors in young people with diagnosed diabetes35,

we considered not only the change in diet scores but also adjusted for the baseline diet quality score in our models. Furthermore, similar to an investigation in the ARIC study25, we have

compared the risk to develop T2D, MI or stroke in categories of increased/decreased diet scores in comparison to a stable reference category. Another approach to consider repeated measures

of diet scores, which was reported in several studies9,11,15, was the calculation of cumulative averages, if at least two time points of dietary measurement were available. Nevertheless,

this would not be the ideal approach to detect changes in dietary intake over time as we intended. The restriction to the two diet quality scores MedPyr and HDS in our study could be a

limitation, since there is current evidence on many more diet quality scores being associated with the cardiometabolic diseases of interest3,25,33. We restricted the analyses to the MedPyr

to follow-up on the previous investigation in our EPIC-Potsdam study: baseline MedPyr adherence was associated with lower T2D risk in the total population and lower MI risk in women7. We

aimed to investigate, if a change towards a higher MedPyr score would be also beneficial in terms of these health outcomes. The focus on the HDS had several reasons: (1) the aforementioned

gap in terms of investigations on disease associations, (2) the consideration of absolute intake ranges comparable to MedPyr, (3) the investigation of a diet score which includes German

dietary recommendations as an alternative approach to a Mediterranean-originated diet. Besides data on dietary intake, several data on socio-demographics and lifestyle factors were collected

with the self-administered questionnaire at FUP3. Hence, we were able to adjust the regression models for several potential confounders on the diet-disease-associations like age, sex,

education and occupation, physical activity, smoking and alcohol consumption and considered both data from baseline and FUP3. Furthermore, we adjusted for self-reported hypertension, since

knowledge about this health condition could have led to a higher awareness for a healthy lifestyle and subsequently to a change in dietary behavior. To rule out that the association was

mainly driven by total energy intake instead of diet composition, the regression models were adjusted for it at baseline and FUP3. We adjusted the models additionally for the baseline diet

quality score to consider the so-called “exogenous change”, which means all non-random changes in the diet quality score at FUP3 which were not predetermined by the diet quality score at

baseline36. Since BMI could theoretically be a confounder or on the mediating path to the chronic diseases of interest, we decided to adjust for BMI at baseline and FUP3 in a separate model.

No major changes in the HRs were observed. Nevertheless, we cannot fully rule out that residual confounding could be still existent. A further strength was the investigation of potential

effect measure modifiers like sex29,30 and baseline diet score. Although the non-significant statistical interactions would not support any sex differences in associations, our observation

of a reduced T2D risk with increased MedPyr might be restricted to women supports sex-stratified analyses. Similar to our observation, Wang et al. also observed stronger associations for the

investigated DQIs in women than in men33. Furthermore, although the statistical interaction with baseline diet score tertiles was not significant, the results observed in the stratified

analyses supported the findings from the main analyses and gave insights beyond that. A further strength was the conduct of more sensitivity analyses, e.g. to eliminate the possibility of

reverse causation. To do so, we excluded those participants, who reported a change in diet due to weight gain, high blood pressure, dyslipidemia or diabetes. Since the results were not

changed, they indicated robust associations. Nevertheless, we could not rule out that there were other unmeasured confounders (e.g. subclinical medical conditions), which could have led to a

change in dietary behavior and were associated with the investigated chronic diseases. CONCLUSION In this middle-aged population of the EPIC-Potsdam study, an increase in adherence to a

Mediterranean Diet, measured with the MedPyr, was associated with a lower T2D risk. While changes in MedPyr were not appreciably related to MI risk, a decrease in MedPyr adherence was

related to higher stroke risk. The findings contribute further evidence on advantages of an actual change in dietary behavior towards a Mediterranean Diet. While higher adherence in HDS at

baseline was related to lower risk of T2D and stroke, a decrease in HDS adherence over time was associated with a higher stroke risk, but changes in HDS were not associated with T2D and MI

risk. DATA AVAILABILITY The datasets generated and/or analysed during the current study are not publicly available due to German Federal and State data protection regulations, but are

available from the secretariat of the Human Study Center ([email protected]) on reasonable request. REFERENCES * Hu, F. B. Dietary pattern analysis: A new direction in nutritional

epidemiology. _Curr. Opin. Lipidol._ 13(1), 3–9 (2002). Article  CAS  PubMed  Google Scholar  * Jannasch, F., Kroger, J. & Schulze, M. B. Dietary patterns and type 2 diabetes: A

systematic literature review and meta-analysis of prospective studies. _J. Nutr._ 147(6), 1174–1182 (2017). Article  CAS  PubMed  Google Scholar  * Schwingshackl, L. & Hoffmann, G. Diet

quality as assessed by the Healthy Eating Index, the Alternate Healthy Eating Index, the Dietary Approaches to Stop Hypertension score, and health outcomes: A systematic review and

meta-analysis of cohort studies. _J. Acad. Nutr. Diet._ 115(5), 780–800 (2015). Article  PubMed  Google Scholar  * Galbete, C., Schwingshackl, L., Schwedhelm, C., Boeing, H. & Schulze,

M. B. Evaluating Mediterranean diet and risk of chronic disease in cohort studies: An umbrella review of meta-analyses. _Eur. J. Epidemiol._ 33(10), 909–931 (2018). Article  PubMed  PubMed

Central  Google Scholar  * Sofi, F., Macchi, C., Abbate, R., Gensini, G. F. & Casini, A. Mediterranean diet and health. _Biofactors_ 39(4), 335–342 (2013). Article  CAS  PubMed  Google

Scholar  * Chiuve, S. E. _et al._ Alternative dietary indices both strongly predict risk of chronic disease. _J. Nutr._ 142(6), 1009–1018 (2012). Article  CAS  PubMed  PubMed Central  Google

Scholar  * Galbete, C. _et al._ Nordic diet, Mediterranean diet, and the risk of chronic diseases: The EPIC-Potsdam study. _BMC Med._ 16(1), 99 (2018). Article  PubMed  PubMed Central 

Google Scholar  * InterAct, C. _et al._ Mediterranean diet and type 2 diabetes risk in the European Prospective Investigation into Cancer and Nutrition (EPIC) study: The InterAct project.

_Diabetes Care_ 34(9), 1913–1918 (2011). Article  Google Scholar  * Tong, T. Y., Wareham, N. J., Khaw, K. T., Imamura, F. & Forouhi, N. G. Prospective association of the Mediterranean

diet with cardiovascular disease incidence and mortality and its population impact in a non-Mediterranean population: The EPIC-Norfolk study. _BMC Med._ 14(1), 135 (2016). Article  PubMed 

PubMed Central  Google Scholar  * Salas-Salvado, J. _et al._ Prevention of diabetes with Mediterranean diets: A subgroup analysis of a randomized trial. _Ann. Intern. Med._ 160(1), 1–10

(2014). Article  PubMed  Google Scholar  * Hu, F. B. _et al._ Dietary fat and coronary heart disease: A comparison of approaches for adjusting for total energy intake and modeling repeated

dietary measurements. _Am. J. Epidemiol._ 149(6), 531–540 (1999). Article  CAS  PubMed  Google Scholar  * Ley, S. H. _et al._ Changes in overall diet quality and subsequent type 2 diabetes

risk: Three U.S. prospective cohorts. _Diabetes Care_ 39(11), 2011–8 (2016). Article  CAS  PubMed  PubMed Central  Google Scholar  * Sotos-Prieto, M. _et al._ Changes in diet quality scores

and risk of cardiovascular disease among US men and women. _Circulation_ 132(23), 2212–2219 (2015). Article  CAS  PubMed  PubMed Central  Google Scholar  * Glenn, A. J. _et al._ Longitudinal

changes in adherence to the portfolio and DASH dietary patterns and cardiometabolic risk factors in the PREDIMED-Plus study. _Clin. Nutr._ 40(5), 2825–2836 (2021). Article  CAS  PubMed 

Google Scholar  * Maddock, J. _et al._ Adherence to a Dietary Approaches to Stop Hypertension (DASH)-type diet over the life course and associated vascular function: A study based on the MRC

1946 British birth cohort. _Br. J. Nutr._ 119(5), 581–589 (2018). Article  CAS  PubMed  PubMed Central  Google Scholar  * Winkvist, A. _et al._ Longitudinal 10-year changes in dietary

intake and associations with cardio-metabolic risk factors in the Northern Sweden Health and Disease Study. _Nutr. J._ 16(1), 20 (2017). Article  PubMed  PubMed Central  Google Scholar  *

Boeing, H., Korfmann, A. & Bergmann, M. M. Recruitment procedures of EPIC-Germany. European investigation into cancer and nutrition. _Ann. Nutr. Metab._ 43(4), 205–15 (1999). Article 

CAS  PubMed  Google Scholar  * Boeing, H., Wahrendorf, J. & Becker, N. EPIC-Germany–A source for studies into diet and risk of chronic diseases. European investigation into cancer and

nutrition. _Ann. Nutr. Metab._ 43(4), 195–204 (1999). Article  CAS  PubMed  Google Scholar  * Bergmann, M. M., Bussas, U. & Boeing, H. Follow-up procedures in EPIC-Germany–data quality

aspects. European prospective investigation into cancer and nutrition. _Ann. Nutr. Metab._ 43(4), 225–34 (1999). Article  CAS  PubMed  Google Scholar  * Nöthlings, U. _Development and

Evaluation of an Abbreviated Food Frequency Questionnaire to Discriminate Between Study Participants in a Cohort Study [Dissertation]_ (Technical University Berlin, 2004). Google Scholar  *

MONICA Mengenliste1991 05.01.2023. file:///C:/Users/Jannasch/Downloads/0654_1271_web_6.pdf. * Trichopoulou, A., Costacou, T., Bamia, C. & Trichopoulos, D. Adherence to a Mediterranean

diet and survival in a Greek population. _N. Engl. J. Med._ 348(26), 2599–2608 (2003). Article  PubMed  Google Scholar  * Jannasch, F., Nickel, D. V., Bergmann, M. M. & Schulze, M. B. A

new evidence-based diet score to capture associations of food consumption and chronic disease risk. _Nutrients_ 14(11), 2359 (2022). Article  CAS  PubMed  PubMed Central  Google Scholar  *

Jannasch, F., Nickel, D. & Schulze, M. B. The reliability and relative validity of predefined dietary patterns were higher than that of exploratory dietary patterns in the European

Prospective Investigation into Cancer and Nutrition (EPIC)-Potsdam population. _Br. J. Nutr._ 125(11), 1270–1280 (2021). Article  CAS  PubMed  Google Scholar  * Xu, Z., Steffen, L. M.,

Selvin, E. & Rebholz, C. M. Diet quality, change in diet quality and risk of incident CVD and diabetes. _Public Health Nutr._ 23(2), 329–338 (2020). Article  PubMed  Google Scholar  *

Bergmann, M. M. _et al._ The association of pattern of lifetime alcohol use and cause of death in the European prospective investigation into cancer and nutrition (EPIC) study. _Int. J.

Epidemiol._ 42(6), 1772–1790 (2013). Article  PubMed  Google Scholar  * Kroke, A. _et al._ Measures of quality control in the German component of the EPIC study. European Prospective

Investigation into Cancer and Nutrition. _Ann. Nutr. Metab._ 43(4), 216–24 (1999). Article  CAS  PubMed  Google Scholar  * Klipstein-Grobusch, K., Georg, T. & Boeing, H. Interviewer

variability in anthropometric measurements and estimates of body composition. _Int. J. Epidemiol._ 26(Suppl 1), S174–S180 (1997). Article  PubMed  Google Scholar  * Meisinger, C. _et al._

Sex differences in risk factors for incident type 2 diabetes mellitus: The MONICA Augsburg cohort study. _Arch. Intern. Med._ 162(1), 82–89 (2002). Article  PubMed  Google Scholar  *

Shufelt, C. L., Pacheco, C., Tweet, M. S. & Miller, V. M. Sex-specific physiology and cardiovascular disease. _Adv. Exp. Med. Biol._ 1065, 433–454 (2018). Article  PubMed  PubMed Central

  Google Scholar  * Romaguera, D. _et al._ Mediterranean diet and type 2 diabetes risk in the European Prospective Investigation into Cancer and Nutrition (EPIC) study: The InterAct project.

_Diabetes Care_ 34(9), 1913–1918 (2011). Article  CAS  PubMed  Google Scholar  * Zhang, C. _et al._ Alcohol intake and risk of stroke: A dose-response meta-analysis of prospective studies.

_Int. J. Cardiol._ 174(3), 669–677 (2014). Article  PubMed  Google Scholar  * Wang, P. _et al._ Optimal dietary patterns for prevention of chronic disease. _Nat. Med._ 29(3), 719–728 (2023).

Article  CAS  PubMed  PubMed Central  Google Scholar  * Lee, D. H. _et al._ Dietary inflammatory and insulinemic potential and risk of type 2 diabetes: Results From three prospective U.S.

cohort studies. _Diabetes Care_ 43(11), 2675–83 (2020). Article  CAS  PubMed  PubMed Central  Google Scholar  * Barnes, T. L. _et al._ Change in DASH diet score and cardiovascular risk

factors in youth with type 1 and type 2 diabetes mellitus: The SEARCH for Diabetes in Youth Study. _Nutr. Diabetes_ 3(10), e91 (2013). Article  CAS  PubMed  PubMed Central  Google Scholar  *

Tennant, P. W. G., Arnold, K. F., Ellison, G. T. H. & Gilthorpe, M. S. Analyses of “change scores” do not estimate causal effects in observational data. _Int. J. Epidemiol._ 51(5),

1604–1615 (2022). Article  PubMed  Google Scholar  Download references ACKNOWLEDGEMENTS We thank the Human Study Centre (HSC) of the German Institute of Human Nutrition Potsdam-Rehbruecke,

namely the trustee and the examination unit for the collection, the data hub for the processing, and the participants for the provision of the data. FUNDING Open Access funding enabled and

organized by Projekt DEAL. The recruitment phase of the EPIC-Potsdam Study was supported by the Federal Ministry of Science, Germany (01 EA 9401) and the European Union (SOC 95201408 05F02).

The follow-up of the EPIC-Potsdam Study was supported by German Cancer Aid (70-2488-Ha I) and the European Community (SOC 98200769 05F02). Furthermore, this work was supported by

NutriAct—Competence Cluster Nutrition Research Berlin-Potsdam funded by the German Federal Ministry of Education and Research (BMBF) (grant number 01EA1408A, 01EA1806A). The publication is

furthermore funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation)—491394008. AUTHOR INFORMATION AUTHORS AND AFFILIATIONS * Department of Molecular Epidemiology,

German Institute of Human Nutrition Potsdam-Rehbruecke, Arthur-Scheunert-Allee 114-116, 14558, Nuthetal, Germany Franziska Jannasch, Daniela V. Nickel, Olga Kuxhaus & Matthias B. Schulze

* NutriAct Competence Cluster for Nutrition Research, Berlin-Potsdam, Germany Franziska Jannasch, Daniela V. Nickel & Matthias B. Schulze * German Center for Diabetes Research (DZD),

München-Neuherberg, Germany Franziska Jannasch, Daniela V. Nickel, Olga Kuxhaus & Matthias B. Schulze * Institute of Nutritional Science, University of Potsdam, Nuthetal, Germany Daniela

V. Nickel & Matthias B. Schulze Authors * Franziska Jannasch View author publications You can also search for this author inPubMed Google Scholar * Daniela V. Nickel View author

publications You can also search for this author inPubMed Google Scholar * Olga Kuxhaus View author publications You can also search for this author inPubMed Google Scholar * Matthias B.

Schulze View author publications You can also search for this author inPubMed Google Scholar CONTRIBUTIONS All authors (F.J., D.V.N., O.K., M.B.S.) contributed to the analysis conception and

design. Data preparation and data analysis were performed by F.J., D.V.N., O.K.. The first draft of the manuscript was written by F.J. and all authors commented on previous versions of the

manuscript. All authors read and approved the final manuscript. CORRESPONDING AUTHOR Correspondence to Franziska Jannasch. ETHICS DECLARATIONS COMPETING INTERESTS The authors declare no

competing interests. ADDITIONAL INFORMATION PUBLISHER'S NOTE Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

SUPPLEMENTARY INFORMATION SUPPLEMENTARY INFORMATION. RIGHTS AND PERMISSIONS OPEN ACCESS This article is licensed under a Creative Commons Attribution 4.0 International License, which permits

use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the

Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless

indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory

regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit

http://creativecommons.org/licenses/by/4.0/. Reprints and permissions ABOUT THIS ARTICLE CITE THIS ARTICLE Jannasch, F., Nickel, D.V., Kuxhaus, O. _et al._ Longitudinally changed diet

quality scores and their association with type 2 diabetes mellitus and cardiovascular diseases in the EPIC-Potsdam study. _Sci Rep_ 14, 13907 (2024).

https://doi.org/10.1038/s41598-024-63899-8 Download citation * Received: 22 November 2023 * Accepted: 03 June 2024 * Published: 17 June 2024 * DOI: https://doi.org/10.1038/s41598-024-63899-8

SHARE THIS ARTICLE Anyone you share the following link with will be able to read this content: Get shareable link Sorry, a shareable link is not currently available for this article. Copy

to clipboard Provided by the Springer Nature SharedIt content-sharing initiative KEYWORDS * Diet quality scores * Mediterranean diet * Type-2-diabetes mellitus * Cardiovascular disease